The Impact of the Quality of Sexual Education on HIV/AIDS in Zambia: Evidence from a 1 Natural Experiment 2

Background: 21 This paper focuses on how sexual education quality is associated with HIV-related knowledge, stigma and 22 risky sexual behaviour among Zambians. The implementation of the Comprehensive Sexuality Education 23 programme since 2014, with sequential in-service teacher training provides a natural experiment for testing the 24 hypothesis that quality of sexual education, not quantity only, is positively associated with HIV knowledge, and 25 negatively associated with risky sexual behaviours and stigma. 26 Methods: 27 Data are drawn from the 2016 Zambian Population-Based HIV Impact Assessment survey, which sampled 28 24,663 individuals aged 15-59 years old nationwide. Province fixed-effects double difference model is used to 29 test our hypothesis. Results: We found that sexual education with well-trained teachers reduced the number of HIV-related stigma by 32 0.13 points, while as expected it was associated with a greater number of correct HIV-related knowledge (0.29 33 points). No significant association was found between quality of sex education and risky sexual behaviours. Conclusion: These results point to the importance of investing in high quality sexual education to combat the HIV-AIDS progression.


5
Skills Education framework (2011) and Comprehensive Sexuality Education framework (2013) have been merged 99 and included in the school curriculum at national level with the aim to reduce the impact of HIV on young people.

100
The CSE is not a standalone subject, but is integrated into other subjects like social studies, civic education, biology, 101 and home economics. The CSE framework covers several topics, including relationship, values, attitudes and skills, 102 culture and human rights, human development, sexual behaviours and sexual and reproductive health. The targeted 103 population are adolescents from grade 5 to 12, with the objective to improve their sexual and reproductive health 104 (SRH) (14). By 2015, the CSE programme had reached 100% of schools, with 77% of learners from grade 5 to 12 105 and 38,251 teachers were trained to teach the CSE curriculum (13).

106
However, in-service teacher training on CSE which prepares teachers to correctly deliver the CSE     In order to estimate the effect of participation in the CSE curriculum on HIV-related knowledge, discriminatory 135 attitudes and risky behaviours, we used a double difference (DD) approach.

136
Because only school age children at the time of the introduction of CSE were exposed to the intervention of 137 improved quality of sexual education, the individual's year of birth or age was used to divide the sample into the 138 Pre-CSE and Post-CSE groups, similar to the before/after treatment concept in impact evaluation studies. In Zambia, 139 school (grade 1) starts at the age of 7, which means that a child reaches grade 5 at age 11 and grade 12 at age 18 (see

148
Although the CSE was implemented in the whole country in 2014, the provision of specific support for in-149 service teacher training was phased. Without adequate training of teachers, the knowledge of the new CSE 150 curriculum is less likely to have been correctly conveyed to students. We therefore have subdivided the provinces

165
In addition to a specific control for education level completed, the X vector includes controls for gender (9), 166 religion, marital status, wealth index, and ethnic background (2,12,21). We also control for HIV status, as it can 167 influence people's HIV-related knowledge, stigma and risky behaviours. Some estimates also control for specific 168 provinces in order to account for regions specificities (more details are given below).

202
Where is the province fixed-effect. It is important to note that because the zone variables are collinear with the 203 province dummies, the estimation controlling for province fixed-effects do not include this variable.

208
In the NB model, the conditional mean of the outcome given the value of predictors is the same as the one for the

219
Third, in order to further test the identification strategy, we constructed a 'placebo' treated group. The

225
All the results presented in this study are weighted to account for the complex sample design and post-226 stratification to control for non-response; variance estimates are calculated using the Jack-knife replicate weights.     Table A1.

234
Differences in outcome variables are examined in detail by cohort and zone of treatment in

265
Turning now to the models estimating the effects of the quality sexual education on HIV-related 266 discriminatory attitudes (Table 4)

12
The robustness checks with the placebo post-CSE cohort presented in Table A2 in Additional file 1 shows 287 no significance of the interaction term coefficients for all the outcomes. This provides confidence that the results 288 obtained are not just driven by the difference that can be found between younger and older cohorts.

304
The findings for the effect of quality sexual education on risky sexual behaviours differed from those for 305 knowledge and attitudes in their absence of an effect of the timing of in-service teacher training for the CSE 306 implementation. The lack of a quality sexual education effect on subsequent sexual behaviour changes may 307 reflect the need for a longer time frame. Behavioural change is a process more than an event and can take many 308 stages (26,27). There were also, in general no effects of level education on risky sexual behaviours.

309
Once again, the results using and an older cohort as placebo for treated group showed non-significant results,

310
providing more confidence to our identification strategy.

311
After controlling for exposure to a quality sexual education program, our study also documents the 312 independent and additional effect of secondary or higher education on HIV-knowledge and reduction of 313 13 discriminatory attitudes. These findings underscore the contributions of years of schooling, which enables 314 individuals to read and process health related information, as well as the content or quality of the health-related 315 materials to which they are exposed in school. These effects are strong for the knowledge and attitudinal 316 domains, and do not carry over into behaviour. As meta-analytic review by Smoak et al. (28), which concluded 317 that HIV-risk interventions neither increased nor decreased sexual occasions and number of sexual partners, our 318 study showed no significant independent effect of additional years of education on risky sexual behaviours.

326
First, we found that the improvement of the quality of sexual education offered by teachers well-trained to 327 deliver the CSE programme increased the number of correct HIV-related knowledge by 0.29 points. Second, the 328 implementation of quality sexual education reduced the number of HIV-related discriminatory attitudes by 0.13 329 points for the exposed group in zones which first received in-service teacher training, compared to the 330 unexposed cohort. Third, no significant effect was found for the number of risky sexual behaviours.

331
There are several limitations to our study. First, because we are not using an experimental design, the 332 identification strategy could be imperfect, notably in terms of the comparability of our exposed and unexposed 333 groups. However, we reduced significantly this potential problem by controlling for observables and non-334 observables specific to each province through province fixed-effect analyses and by reducing the age range of

14
Nevertheless, this study contributes to the literature in many ways. By taking advantage of the introduction 341 of a better-quality sexual education with sequential in-service teacher training in 2014 for a natural experiment,

342
we were able to split the effect of education into quality versus quantity of education. Our study shows the 343 importance of considering the quality of sexual education when assessing the relation between education and

344
HIV-related knowledge and behaviours. Second, the study analyses the effect of quality sex education on both

345
HIV-related knowledge, discriminatory attitudes and risky sexual behaviours using additive indexes for each

347
With this specificity, the study provides actionable results to those seeking to further reduce HIV incidence

380
The risky behaviour estimations are based on respondents who ever had sexual intercourse.

385
Competing Interests: The authors declare that they have no conflict of interest.